Journal
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
Volume 43, Issue 10, Pages 3432-3445Publisher
IEEE COMPUTER SOC
DOI: 10.1109/TPAMI.2020.2988262
Keywords
Convolutional neural networks; Quantization (signal); Learning (artificial intelligence); Noise reduction; Machine learning; Convolution; Binary convolutional neural networks; channel-wise interactions; deep reinforcement learning; hierarchical reinforcement learning; feature denoising
Funding
- National Key Research and Development Program of China [2017YFA0700802]
- National Natural Science Foundation of China [61822603, U1813218, U1713214, 61672306]
- Beijing Academy of Artificial Intelligence (BAAI) [BAAI2020ZJ0202]
- Institute for Guo Qiang, Tsinghua University
- Shenzhen Fundamental Research Fund [JCYJ20170412170602564]
- Tsinghua University Initiative Scientific Research Program
Ask authors/readers for more resources
This paper proposes a channel-wise interaction based binary convolutional neural networks (CI-BCNN) approach for efficient inference. By using reinforcement learning to mine channel-wise interactions, correct inconsistent signs, and alleviate noise in channel-wise priors, the proposed approach improves inference efficiency.
In this paper, we propose a channel-wise interaction based binary convolutional neural networks (CI-BCNN) approach for efficient inference. Conventional binary convolutional neural networks usually apply the xnor and bitcount operations in the binary convolution with notable quantization errors, which obtain opposite signs of pixels in binary feature maps compared to their full-precision counterparts and lead to significant information loss. In our proposed CI-BCNN method, we exploit the channel-wise interactions with the prior knowledge which aims to alleviate inconsistency of signs in binary feature maps and preserves the information of input samples during inference. Specifically, we mine the channel-wise interactions by using a reinforcement learning model, and impose channel-wise priors on the intermediate feature maps to correct inconsistent signs through the interacted bitcount. Since CI-BCNN mines the channel-wise interactions in a large search space where each channel may correlate with others, the search deficiency caused by sparse interactions obstacles the agent to obtain the optimal policy. To address this, we further present a hierarchical channel-wise interaction based binary convolutional neural networks (HCI-BCNN) method to shrink the search space via hierarchical reinforcement learning. Moreover, we propose a denoising interacted bitcount operation in binary convolution by smoothing the channel-wise interactions, so that noise in channel-wise priors can be alleviated. Extensive experimental results on the CIFAR-10 and ImageNet datasets demonstrate the effectiveness of the proposed CI-BCNN and HCI-BCNN.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available